I have read an article talking about binary clustering using Matrix factorization(see attached), but i would like to understand some optimization concepts:
- Is it reasonable to use a Frobenius norm in such optimization?
- What does it mean the centroid constraint: C^T*1=0, and why it is equivalent to K-means especially in this case (When rho is large)?
- Is there other constraints that could improve clustering optimization?
- Is there other optimization technique in place of DPLM(Discrete Proximal Linearized Minimization)?